OT on the Map: Quantifying Domain Shifts in Geographic Space

TL;DR

This paper introduces GeoSpOT, a method leveraging geographic information and Optimal Transport to quantify domain shifts in geospatial data, aiding in predicting cross-region model transfer success.

Contribution

It proposes a novel distance measure for geospatial domains using Optimal Transport and demonstrates its effectiveness in predicting transfer difficulty and guiding data selection.

Findings

GeoSpOT distances predict cross-domain transfer difficulty effectively.

Pretrained location embeddings provide comparable information to image/text embeddings.

GeoSpOT enables preemptive analysis of regions where models may underperform.

Abstract

In computer vision and machine learning for geographic data, out-of-domain generalization is a pervasive challenge, arising from uneven global data coverage and distribution shifts across geographic regions. Though models are frequently trained in one region and deployed in another, there is no principled method for determining when this cross-region adaptation will be successful. A well-defined notion of distance between distributions can effectively quantify how different a new target domain is compared to the domains used for model training, which in turn could support model training and deployment decisions. In this paper, we propose a strategy for computing distances between geospatial domains that leverages geographic information with Optimal Transport methods (GeoSpOT). In our experiments, GeoSpOT distances emerge as effective predictors of cross-domain transfer difficulty. We further demonstrate that embeddings from pretrained location encoders provide information comparable to image/text embeddings, despite relying solely on longitude-latitude pairs as input. This allows users to get an approximation of out-of-domain performance for geospatial models, even when the exact downstream task is unknown, or no task-specific data is available. Building on these findings, we show that GeoSpOT distances can preemptively guide data selection and enable predictive tools to analyze regions where a model is likely to underperform.